I wanted to prove the work energy theorem in three dimensions starting from Newton's second law of motion. I am having some trouble understanding differential swapping and deriving the kinetic energy formula in three dimensions. Assuming $F_c$ is the sum of all conservative force and $F_{nc}$ is a non-conservative force, for one dimension I have the following.


$$\int_{x_{1}}^{x_{2}}F_{c}\ dx\ +\ \int_{x_{1}}^{x_{2}}F_{nc}\ dx\ =m\int_{x_{1}}^{x_{2}}adx$$

$$adx=\frac{dv}{dt}dx=\frac{dx}{dt}dv=vdv \; \label{*} \tag{*}$$

$$m\int_{v\left(x_{1}\right)}^{v\left(x_{2}\right)}vdv\ =\ \frac{1}{2}mv\left(x_{2}\right)^{2}-\frac{1}{2}mv\left(x_{1}\right)^{2}=KE_f-KE_i$$ Using $\frac{dU}{dx}=-F_{c}$, and assuming integrating $F_{nc}$ gives a change in energy $ΔW_{nc}$, we have, $$U_{i}-U_{f}\ +\ ΔW_{nc}=KE_{f}-KE_{i}.$$

In the manipulation to obtain equation $\eqref{*}$, why exactly are we allowed to swap the differentials? Is there a rigorous way to show why you're allowed to do that?

Moreover for the 3D case, assuming that an object is moving along a curve $C$, $$\vec{F_c}+\vec{F_{nc}}=m\vec{a},$$ $$\int_C \vec{F_c}\cdot d\vec{r} +\int_C \vec{F_{nc}}\cdot d\vec{r} = m\int_C \vec{a}\cdot d\vec{r},$$ and further using, $$\nabla U=-\vec{F_c}$$ $$\int_C \vec{F_c}\cdot d\vec{r}=-\int_C\nabla U\cdot d\vec{r} = U_i-U_f,$$we have $$U_{i}-U_{f}\ +\ ΔW_{nc} = m\int_C (a_x \hat{i}_x + a_y \hat{i}_y + a_z \hat{i}_z) \cdot d\vec{r}.$$

How exactly do we handle the final integral on the right? I’m sort of confused since $a$ is an acceleration field now, and I’m not sure if I can do the differential swap trick from before. Is $\vec{a}$ the time derivative of a velocity field now? Also on a different note, could that integral be evaluated using the fundamental theorem of line integrals?

The final result $KE_f-KE_i$ involves computing a value at the beginning and end of the curve (similar to what FTC of line integrals does). That's why I was asking. Thanks!

  • $\begingroup$ I haven't looked at your calculations but why do you think this is the conservation of mechanical energy? Mechanical energy is not conserved when you have non-conservative forces. $\endgroup$ – Dvij D.C. Mar 17 at 5:52
  • $\begingroup$ @DvijD.C. The general formula for conservation of energy I've seen is $U_i+KE_i+ΔW = U_f + KE_f$. I think without the $ΔW$ term it would truly be conservation of mech. energy, but I included a non-conservative force to get the $ΔW$ term in the final equation. If there's some nuance I'm missing I could edit it out. The non-conservative force isn't important to my question. $\endgroup$ – Byte _ Mar 17 at 6:10
  • $\begingroup$ Yes, you can derive the work-energy theorem but it is not the conservation of energy. What you are writing is not the general formula for conservation of energy, conservation of energy is precisely when that $\Delta W_{nc}$ is zero. What you are saying the equivalent of saying "being an omnivore is a generalization of being a vegan", it is just wrong. :) $\endgroup$ – Dvij D.C. Mar 17 at 6:19
  • 2
    $\begingroup$ Oh ok, that makes more sense. I guess I am specifically deriving the Work energy theorem then, not conservation of energy. I'll edit the post $\endgroup$ – Byte _ Mar 17 at 6:24

Thanks to the OP for posting this great question. Indeed in mathematical physics it is more expedient to manipulate integrals and derivatives using the approach of the infinitesimal calculus or the nonstandard calculus. However, mathematical rigor can often provide clarity and avoid differentiability issues which lead to erroneous calculations.

Addressing the first part of the question requires application of the change of variables theorem. We denote the twice differentiable $C^2(\mathbb{R})$ (for convenience) scalar functions $x: t \mapsto x(t), v: t \mapsto \frac{dx}{dt}(t)$, $a: t \mapsto \frac{d^2x}{dt^2}(t)$ and $F_c + F_{nc}: x \mapsto F_c(x) + F_{nc}(x)$. The definite integral manipulation corresponding to the mathematically accurate definition of the mass normalized work ($\frac{1}{m} \int_{x_1}^{x_2} (F_c + F_{nc})(x) \; dx := \int_{t_1}^{t_2} a(x(t)) \frac{dx}{dt}(t) \; dt$), which uses the change of variable $x = x(t)$, given as $$\frac{1}{m} \int_{x_1}^{x_2} (F_c + F_{nc})(x) \; dx = \int_{x_1}^{x_2} a(x) \; dx = \int_{t_1}^{t_2} a(x(t)) \frac{dx}{dt}(t) \; dt,$$ where $x(t_1) = x_1$ and $x(t_2) = x_2$ implies that $$\int_{x_1}^{x_2} a(x) \; dx = \int_{t_1}^{t_2} \frac{dv}{dt}(t) v(t) \; dt,$$ while the change of variable $v = v(t)$ (applied in reverse, to state the manipulation in words) implies that $$\int_{x_1}^{x_2} a(x) \; dx = \int_{t_1}^{t_2} \frac{dv}{dt}(t) v(t) \; dt = \int_{t_1}^{t_2} v(v(t)) \frac{dv}{dt}(t) \; dt = \int_{v_1}^{v_2} v(v) \; dv = {\Huge[} \frac{v^2}{2}{\Huge]}{\Huge|}_{v_1}^{v_2},$$ where $v(t_1) = v_1$ and $v(t_2) = v_2$. The work-energy theorem in the one dimensional case follows directly from this result. This completes the first part of the answer.

The vector dot product $(a_x \hat{i}_{x} + a_y \hat{i}_{y} + a_z \hat{i}_{z}) \cdot d\vec{r} = a_x d x + a_y dy + a_z dz$ so that, using the previous result, $$\int_{C} (a_x \hat{x} + a_y \hat{y} + a_z \hat{z}) \cdot d\vec{r} = \int_{(x_1,y_1,z_1)}^{(x_2,y_2,z_2)} (a_x \hat{x} + a_y \hat{y} + a_z \hat{z}) \cdot d\vec{r} \\ = \int_{x_1}^{x_2} a_x dx + \int_{y_1}^{y_2} a_y dy + \int_{z_1}^{z_2} a_z dz = \frac{1}{2}({\large[}v_x^2{\large]}{\Huge|}_{v_{x1}}^{v_{x2}} + {\large[}v_y^2{\large]}{\Huge|}_{v_{y1}}^{v_{y2}} + {\large[}v_z^2{\large]}){\Huge|}_{v_{z1}}^{v_{z2}} = \frac{1}{2} [\|\vec{v}\|^2]{\Huge|}_{\vec{v}_1}^{\vec{v}_2},$$ which completes the second part of the answer.

Finally, defining $\Delta W_{c} := \int_{x_1}^{x_2} \vec{F}_{c} \cdot d \vec{r} := \int_{x_1}^{x_2} \vec{F}_{c} \cdot \frac{d \vec{v}}{dt}(t) \; dt$, $\nabla V(\vec{r}) := - \vec{F}_c(\vec{r})$ (such a potential function $V \in C^1(\mathbb{R})$ exists due to the definition of conservative forces (which are not necessarily central forces but have a form related to the central forces) $\vec{F}_c(\vec{r}) := \|\vec{F}_c (\vec{r}) \| \frac{\vec{r}}{\|\vec{r}\|}$), $\Delta W_{nc} := \int_{x_1}^{x_2} \vec{F}_{nc} \cdot d \vec{r} := \int_{x_1}^{x_2} \vec{F}_{nc} \cdot \frac{d \vec{v}}{dt}(t) \; dt$ and $T(\vec{v}) := \frac{1}{2} m \|\vec{v}\|^2$, we can state the work-energy theorem for a material particle on which conservative and non-conservative forces act as $$\Delta W_{c} + \Delta W_{nc} = V(\vec{r}_1) - V(\vec{r}_2) + \Delta W_{nc} = T(\vec{v}_2) - T(\vec{v}_1),$$ with the conservation form given as $$V(\vec{r}_1) + T(\vec{v}_1) = V(\vec{r}_2) + T(\vec{v}_2) - \Delta W_{nc}.$$.

  • 1
    $\begingroup$ Thank you. I didn't really like the idea of seeing the differentials 'cancel out', so seeing this answered with the change of variables theorem concretely is nice. I think I was missing the substitution $v=v(t)$. It's simple, but doing that gets rid of time (cancels the dt differential rigorously) and that's what I wanted. It feels sort of weird that you can integrate with respect to $dv_x$, $dv_y$, and $dv_z$ for the multivariable case, though I could get used to it $\endgroup$ – Byte _ Mar 17 at 17:27
  • $\begingroup$ @Byte_, indeed it is more expedient in mathematical physics to manipulate integrals and derivatives using the approach of infinitesimal calculus or nonstandard calculus. However, mathematical rigor can often provide clarity and avoid differentiability issues which lead to erroneous calculations. $\endgroup$ – kbakshi314 Mar 17 at 18:53

If you remember the definition of a differential then you'd know that you can write $dx$ as $$dx = \frac{dx}{dt}dt$$ So that you get $$\frac{dv}{dt}dx = \frac{dv}{dt} \frac{dx}{dt} dt$$ $$ = \left( \frac{dv}{dt} dt \right) \frac{dx}{dt}$$ $$ = \frac{dx}{dt} dv$$ As for the 3d case, you already did half the work. Write $\mathbf{dr}$ as $dx \hat{i} + dy \hat{j} + dz \hat{k}$ and finish the dot product. Then do the same thing component wise what you did in the 1d case you'll get $$\frac{1}{2} m v_x^2 + \frac{1}{2} m v_y^2 + \frac{1}{2} m v_z^2$$

  • $\begingroup$ Where exactly did you get that definition of a differential from? I could sort of see how the 'dt' cancels out, but I thought you could only do that type of substitution using change of variables $\endgroup$ – Byte _ Mar 17 at 17:36
  • $\begingroup$ @Byte_ en.m.wikipedia.org/wiki/Differential_of_a_function $\endgroup$ – Brain Stroke Patient Mar 18 at 19:38
  • $\begingroup$ @Byte_ This has nothing to do with $dt$ cancelling out btw. It's just how differentials are formally defined. $\endgroup$ – Brain Stroke Patient Mar 19 at 9:01

First of all,mechanical energy isn't conserved when $\vec{F_{nc}}$ is there in the system unless it does zero work.So the proof is coming not exactly similar to proof of work energy theorem but quite relatable to that.

Coming to swapping differentials, isn't that like this: $$adx=\frac{dv}{dt}dx$$ Multiplying and dividing by $dt$ $$adx=\frac{dv}{dt}\frac{dx}{dt}dt$$ $$ = \left( \frac{dv}{dt} dt \right) \frac{dx}{dt}$$ $$ = \frac{dx}{dt} dv$$ Now, $$adx=vdv$$

The last integral $\int_C (a_x \hat{i}_x + a_y \hat{i}_y + a_z \hat{i}_z) \cdot d\vec{r}$ can be carried out if $d\vec{r}$ is written as: $$dx\hat{i}_x+dy\hat{i}_y+dz \hat{i}_z$$ Carrying out the dot product with $\vec{a}$, it will be given as: $$a_{x}dx+a_{y}dy+a_{z}dz$$

It's easy to swap differential here as we get $a_{x}dx=v_{x}dv_{x}$, $a_{y}dy=v_{y}dv_{y}$ and $a_{z}dx=v_{z}dv_{z}$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.